Deciphering octopus arm and sucker control mechanisms for novel soft robotic applications

Soft robotics has gained substantial popularity over the last ten years. As the name implies, these robots are mainly composed of soft materials. However, their biggest strength is also their biggest challenge: without rigid structures, movements can be performed with substantially more degrees of freedom than in regular robots. This large motion repertoire makes it hard to develop systems that can reliably predict and control movements. In nature, one model facing similar challenges is the octopus. Octopuses are among the most intelligent molluscs and have developed unique abilities to control all eight of their soft limbs. Each arm is a muscular hydrostat that creates movement through the local interplay of antagonistic muscles and connective tissues. This enables octopuses to perform motions like extension, shortening, bending and stiffening at any location along the arm. Although the arm can essentially move with infinite degrees of freedom, several octopus arm motions, such as extension and fetch, are performed following stereotypical patterns, suggesting that control can be simplified through distributed motor programmes. However, a cause-and-effect understanding linking muscle recruitment, deformation and behaviour remains incomplete. This thesis investigates octopus arm and sucker control across three levels: (1) local motor control underlying stereotyped arm deformations, (2) role of sensory input at the sucker level in CNS-dependent action selection during retrieval, and (3) stereotyped sucker attachment and detachment strategies and their associated pressure dynamics. To address this, we combined ex vivo electromyography with kinematic tracking and muscle deformation measurements, behavioural and movement assays in vivo and ex vivo, and video- and pressure-based analyses of sucker attachment. Our results show patterns of longitudinal and transverse muscle co-activation during electrically evoked arm extension and mechanically induced withdrawal-like resistance, consistent with a stabilising role of the proximal arm region and compatible with local stiffening. Muscle deformation measurements further indicate that neural activity can shape how strain is distributed during pulling. In vivo, octopuses adjusted retrieval decisions and kinematics based on stimulus value: Larger food items likely engage more chemo- and mechanotactile receptors at the sucker-prey contact site and were associated with faster initiation and higher fetch velocities. Non-food items were unlikely to be retrieved, suggesting that the sucker chemosensory system contributes to brain action selection. Consistent with this, isolated arms did not differentiate between food and non-food items, indicating that chemosensory cues at the sucker level bias action selection primarily when central processing is available. Finally, we identified stereotyped patterns of sucker attachment and detachment. Moreover, pressure traces revealed strategy-dependent dynamics, including stepwise reinforcement, plateaus, and gradual versus abrupt release, consistent with active modulation of suction during attachment and detachment. By linking local muscle recruitment, sucker sensory-driven action selection and sucker attachment dynamics, this thesis provides biological principles for reducing control complexity in compliant systems and may inform soft robotic designs that integrate distributed sensing with robust, damage-minimising attachment.